Staged Training Report ✓ Complete

Run ID: shoulder_session_multiheight_large_decoder
Generated: 2026-02-23 04:03:46
Stages Completed: 1
Total Elapsed Time: 06:09:43

Configuration

Config Defaults Changed Since Last Commit

Parameter	Previous	Current
`batch_size`	1	8
`plateau_sweep.plateau_ema_alpha`	0.9	0.85
`plateau_sweep.plateau_improvement_threshold`	0.0005	0.0015

All Configuration Parameters (60 parameters)

Parameter	Value
`total_samples`	10000000
`batch_size`	8
`stage_samples_multiplier`	100000000000
`update_interval`	250
`window_size`	100
`num_best_models_to_keep`	1
`sampling_mode`	Loss-weighted
`loss_weight_temperature`	0.5
`loss_weight_refresh_interval`	50
`stop_on_divergence`	True
`divergence_gap`	0.002
`divergence_ratio`	1.5
`divergence_patience`	50
`divergence_min_updates`	10
`val_spike_threshold`	2.0
`val_spike_window`	15
`val_spike_frequency`	0.75
`val_plateau_patience`	250
`val_plateau_min_delta`	0.0001
`custom_lr`	0.0001
`disable_lr_scaling`	True
`custom_warmup`	-1
`lr_min_ratio`	0.001
`resume_warmup_ratio`	0.05
`plateau_factor`	0.8
`plateau_patience`	15
`preserve_optimizer`	False
`preserve_scheduler`	True
`samples_mode`	Train additional samples
`num_random_obs_to_visualize`	2
`selected_frame_offset`	3
`runs_per_stage`	5
`serial_runs`	True
`clean_old_checkpoints`	True
`enable_baseline`	False
`baseline_runs_per_stage`	1
`run_id`	shoulder_session_multiheight_large_decoder
`enable_wandb`	True
`wandb_project`	developmental-robot-movement
`lr_sweep.lr_min`	1e-07
`lr_sweep.lr_max`	0.01
`lr_sweep.phase_a_num_candidates`	5
`lr_sweep.phase_a_seeds`	1
`lr_sweep.phase_a_time_budget_min`	3.0
`lr_sweep.phase_a_survivor_count`	2
`lr_sweep.phase_b_seeds`	3
`lr_sweep.phase_b_time_budget_min`	10.0
`lr_sweep.ranking_metric`	median_best_val
`lr_sweep.min_samples_before_timeout`	1000
`lr_sweep.min_evals_before_stop`	5
`lr_sweep.save_sweep_state`	True
`plateau_sweep.enabled`	True
`plateau_sweep.plateau_ema_alpha`	0.85
`plateau_sweep.plateau_improvement_threshold`	0.0015
`plateau_sweep.plateau_patience`	25
`plateau_sweep.cooldown_updates`	5
`plateau_sweep.max_sweeps_per_stage`	2
`plateau_sweep.min_sweep_improvement`	0.0
`initial_sweep_enabled`	True
`stage_time_budget_min`	180

Timing Summary

Stage	Plateau Sweeps	Sweep Time	Training Time	Stage Total
Stage 1	5	01:18:45	00:12:38	01:31:24
TOTAL	5	01:18:45	00:12:38	01:31:24

Initial LR Sweep: Stage 1: selected LR 3.16e-05 in 00:15:08

Plateau Sweep Details

Total Sweeps: 5
Stages with Sweeps: 1 of 1
Total Sweep Time: 01:18:45
Average Sweep Duration: 00:15:45

Stage 1: 5 sweeps

LR Progression: 3.2e-05 → 3.2e-05 → 3.2e-05 → 3.2e-05 → 1.8e-06 → 1.8e-06

Sweep #	Triggered At (samples)	Wall Time	Selected LR	Duration
1	16,896	00:02:48	3.16e-05	00:14:07
2	26,112	00:18:26	3.16e-05	00:16:14
3	33,280	00:35:52	3.16e-05	00:16:02
4	55,040	00:55:27	1.78e-06	00:16:09
5	67,584	01:13:41	1.78e-06	00:16:12

Stage Results

Stage	Best Loss	Stop Reason	Samples Trained	Time	Sweeps	LR (Initial→Final)
Stage 1	0.041913	max_sweeps (2)	8,960	01:31:24	5	3.2e-05→1.8e-06

Total Plateau Sweeps: 5

Stop Reason Breakdown

max_sweeps: 1 stages

Loss Across Full Training Run

Loss Detail (Post Initial Drop)

Multi-Run Statistics

Total Runs: 5
Average Best Loss: 0.068729 ± 0.031378
Best Overall: 0.041913
Worst Overall: 0.114280

Stage 1 (5 runs)

Run	Best Loss	Stop Reason	Samples	Time	Selected
1	0.043195	max_sweeps (2)	6,912	01:25:33
2	0.041913	max_sweeps (2)	8,960	01:31:24	✓
3	0.045168	max_sweeps (2)	6,656	01:41:30
4	0.099091	max_sweeps (2)	12,544	00:37:22
5	0.114280	max_sweeps (2)	9,984	00:36:14
Mean: 0.068729 ± 0.031378		Min: 0.041913 / Max: 0.114280		Range: 0.072367

Best Checkpoint

Name: best_model_auto_session_so101_multiheight_part1_1345_shoulder_session_multiheight_large_decoder_00070144_cont_val_0.041913.pth
Stage: 1
Hybrid Loss (full session): 0.051272

Learning Rate Timeline with Plateau Sweeps

Stage Progression

Stage	Orig Loss	Train Loss	Time	Samples	Stop Reason
1 ⭐	0.051272	0.041913	01:31:24	8960	max_sweeps (2)

Hybrid Loss Over Original Session (per Stage)

Stage 1 (Best) - Hybrid Loss: 0.051272

Sample Counts

Cumulative Across All Stages

Per Stage

Stage 1 (Best) - Total Samples: 8,960

Best Checkpoint Inference

Selected Frame 3

Action 0

Action 1

Action 2

Random Observations

Observation 925

Action 0

Action 1

Action 2

Observation 164

Action 0

Action 1

Action 2

Individual Stage Reports

Stage 1 Run 2 (Staged) (orig loss: 0.051272) ⭐